includes/Sanitizer.php

   1 <?php
   2 /**
   3  * XHTML sanitizer for MediaWiki
   4  *
   5  * Copyright (C) 2002-2005 Brion Vibber <brion@pobox.com> et al
   6  * http://www.mediawiki.org/
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License as published by
  10  * the Free Software Foundation; either version 2 of the License, or
  11  * (at your option) any later version.
  12  *
  13  * This program is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  16  * GNU General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU General Public License along
  19  * with this program; if not, write to the Free Software Foundation, Inc.,
  20  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  21  * http://www.gnu.org/copyleft/gpl.html
  22  *
  23  * @file
  24  * @ingroup Parser
  25  */
  26
  27 /**
  28  * Regular expression to match various types of character references in
  29  * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
  30  */
  31 define( 'MW_CHAR_REFS_REGEX',
  32         '/&([A-Za-z0-9\x80-\xff]+);
  33          |&\#([0-9]+);
  34          |&\#x([0-9A-Za-z]+);
  35          |&\#X([0-9A-Za-z]+);
  36          |(&)/x' );
  37
  38 /**
  39  * Regular expression to match HTML/XML attribute pairs within a tag.
  40  * Allows some... latitude.
  41  * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
  42  */
  43 $attrib = '[A-Za-z0-9]';
  44 $space = '[\x09\x0a\x0d\x20]';
  45 define( 'MW_ATTRIBS_REGEX',
  46         "/(?:^|$space)($attrib+)
  47           ($space*=$space*
  48                 (?:
  49                  # The attribute value: quoted or alone
  50                   \"([^<\"]*)\"
  51                  | '([^<']*)'
  52                  |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
  53                  |  (\#[0-9a-fA-F]+) # Technically wrong, but lots of
  54                                                          # colors are specified like this.
  55                                                          # We'll be normalizing it.
  56                 )
  57            )?(?=$space|\$)/sx" );
  58
  59 /**
  60  * List of all named character entities defined in HTML 4.01
  61  * http://www.w3.org/TR/html4/sgml/entities.html
  62  * @private
  63  */
  64 global $wgHtmlEntities;
  65 $wgHtmlEntities = array(
  66         'Aacute'   => 193,
  67         'aacute'   => 225,
  68         'Acirc'    => 194,
  69         'acirc'    => 226,
  70         'acute'    => 180,
  71         'AElig'    => 198,
  72         'aelig'    => 230,
  73         'Agrave'   => 192,
  74         'agrave'   => 224,
  75         'alefsym'  => 8501,
  76         'Alpha'    => 913,
  77         'alpha'    => 945,
  78         'amp'      => 38,
  79         'and'      => 8743,
  80         'ang'      => 8736,
  81         'Aring'    => 197,
  82         'aring'    => 229,
  83         'asymp'    => 8776,
  84         'Atilde'   => 195,
  85         'atilde'   => 227,
  86         'Auml'     => 196,
  87         'auml'     => 228,
  88         'bdquo'    => 8222,
  89         'Beta'     => 914,
  90         'beta'     => 946,
  91         'brvbar'   => 166,
  92         'bull'     => 8226,
  93         'cap'      => 8745,
  94         'Ccedil'   => 199,
  95         'ccedil'   => 231,
  96         'cedil'    => 184,
  97         'cent'     => 162,
  98         'Chi'      => 935,
  99         'chi'      => 967,
 100         'circ'     => 710,
 101         'clubs'    => 9827,
 102         'cong'     => 8773,
 103         'copy'     => 169,
 104         'crarr'    => 8629,
 105         'cup'      => 8746,
 106         'curren'   => 164,
 107         'dagger'   => 8224,
 108         'Dagger'   => 8225,
 109         'darr'     => 8595,
 110         'dArr'     => 8659,
 111         'deg'      => 176,
 112         'Delta'    => 916,
 113         'delta'    => 948,
 114         'diams'    => 9830,
 115         'divide'   => 247,
 116         'Eacute'   => 201,
 117         'eacute'   => 233,
 118         'Ecirc'    => 202,
 119         'ecirc'    => 234,
 120         'Egrave'   => 200,
 121         'egrave'   => 232,
 122         'empty'    => 8709,
 123         'emsp'     => 8195,
 124         'ensp'     => 8194,
 125         'Epsilon'  => 917,
 126         'epsilon'  => 949,
 127         'equiv'    => 8801,
 128         'Eta'      => 919,
 129         'eta'      => 951,
 130         'ETH'      => 208,
 131         'eth'      => 240,
 132         'Euml'     => 203,
 133         'euml'     => 235,
 134         'euro'     => 8364,
 135         'exist'    => 8707,
 136         'fnof'     => 402,
 137         'forall'   => 8704,
 138         'frac12'   => 189,
 139         'frac14'   => 188,
 140         'frac34'   => 190,
 141         'frasl'    => 8260,
 142         'Gamma'    => 915,
 143         'gamma'    => 947,
 144         'ge'       => 8805,
 145         'gt'       => 62,
 146         'harr'     => 8596,
 147         'hArr'     => 8660,
 148         'hearts'   => 9829,
 149         'hellip'   => 8230,
 150         'Iacute'   => 205,
 151         'iacute'   => 237,
 152         'Icirc'    => 206,
 153         'icirc'    => 238,
 154         'iexcl'    => 161,
 155         'Igrave'   => 204,
 156         'igrave'   => 236,
 157         'image'    => 8465,
 158         'infin'    => 8734,
 159         'int'      => 8747,
 160         'Iota'     => 921,
 161         'iota'     => 953,
 162         'iquest'   => 191,
 163         'isin'     => 8712,
 164         'Iuml'     => 207,
 165         'iuml'     => 239,
 166         'Kappa'    => 922,
 167         'kappa'    => 954,
 168         'Lambda'   => 923,
 169         'lambda'   => 955,
 170         'lang'     => 9001,
 171         'laquo'    => 171,
 172         'larr'     => 8592,
 173         'lArr'     => 8656,
 174         'lceil'    => 8968,
 175         'ldquo'    => 8220,
 176         'le'       => 8804,
 177         'lfloor'   => 8970,
 178         'lowast'   => 8727,
 179         'loz'      => 9674,
 180         'lrm'      => 8206,
 181         'lsaquo'   => 8249,
 182         'lsquo'    => 8216,
 183         'lt'       => 60,
 184         'macr'     => 175,
 185         'mdash'    => 8212,
 186         'micro'    => 181,
 187         'middot'   => 183,
 188         'minus'    => 8722,
 189         'Mu'       => 924,
 190         'mu'       => 956,
 191         'nabla'    => 8711,
 192         'nbsp'     => 160,
 193         'ndash'    => 8211,
 194         'ne'       => 8800,
 195         'ni'       => 8715,
 196         'not'      => 172,
 197         'notin'    => 8713,
 198         'nsub'     => 8836,
 199         'Ntilde'   => 209,
 200         'ntilde'   => 241,
 201         'Nu'       => 925,
 202         'nu'       => 957,
 203         'Oacute'   => 211,
 204         'oacute'   => 243,
 205         'Ocirc'    => 212,
 206         'ocirc'    => 244,
 207         'OElig'    => 338,
 208         'oelig'    => 339,
 209         'Ograve'   => 210,
 210         'ograve'   => 242,
 211         'oline'    => 8254,
 212         'Omega'    => 937,
 213         'omega'    => 969,
 214         'Omicron'  => 927,
 215         'omicron'  => 959,
 216         'oplus'    => 8853,
 217         'or'       => 8744,
 218         'ordf'     => 170,
 219         'ordm'     => 186,
 220         'Oslash'   => 216,
 221         'oslash'   => 248,
 222         'Otilde'   => 213,
 223         'otilde'   => 245,
 224         'otimes'   => 8855,
 225         'Ouml'     => 214,
 226         'ouml'     => 246,
 227         'para'     => 182,
 228         'part'     => 8706,
 229         'permil'   => 8240,
 230         'perp'     => 8869,
 231         'Phi'      => 934,
 232         'phi'      => 966,
 233         'Pi'       => 928,
 234         'pi'       => 960,
 235         'piv'      => 982,
 236         'plusmn'   => 177,
 237         'pound'    => 163,
 238         'prime'    => 8242,
 239         'Prime'    => 8243,
 240         'prod'     => 8719,
 241         'prop'     => 8733,
 242         'Psi'      => 936,
 243         'psi'      => 968,
 244         'quot'     => 34,
 245         'radic'    => 8730,
 246         'rang'     => 9002,
 247         'raquo'    => 187,
 248         'rarr'     => 8594,
 249         'rArr'     => 8658,
 250         'rceil'    => 8969,
 251         'rdquo'    => 8221,
 252         'real'     => 8476,
 253         'reg'      => 174,
 254         'rfloor'   => 8971,
 255         'Rho'      => 929,
 256         'rho'      => 961,
 257         'rlm'      => 8207,
 258         'rsaquo'   => 8250,
 259         'rsquo'    => 8217,
 260         'sbquo'    => 8218,
 261         'Scaron'   => 352,
 262         'scaron'   => 353,
 263         'sdot'     => 8901,
 264         'sect'     => 167,
 265         'shy'      => 173,
 266         'Sigma'    => 931,
 267         'sigma'    => 963,
 268         'sigmaf'   => 962,
 269         'sim'      => 8764,
 270         'spades'   => 9824,
 271         'sub'      => 8834,
 272         'sube'     => 8838,
 273         'sum'      => 8721,
 274         'sup'      => 8835,
 275         'sup1'     => 185,
 276         'sup2'     => 178,
 277         'sup3'     => 179,
 278         'supe'     => 8839,
 279         'szlig'    => 223,
 280         'Tau'      => 932,
 281         'tau'      => 964,
 282         'there4'   => 8756,
 283         'Theta'    => 920,
 284         'theta'    => 952,
 285         'thetasym' => 977,
 286         'thinsp'   => 8201,
 287         'THORN'    => 222,
 288         'thorn'    => 254,
 289         'tilde'    => 732,
 290         'times'    => 215,
 291         'trade'    => 8482,
 292         'Uacute'   => 218,
 293         'uacute'   => 250,
 294         'uarr'     => 8593,
 295         'uArr'     => 8657,
 296         'Ucirc'    => 219,
 297         'ucirc'    => 251,
 298         'Ugrave'   => 217,
 299         'ugrave'   => 249,
 300         'uml'      => 168,
 301         'upsih'    => 978,
 302         'Upsilon'  => 933,
 303         'upsilon'  => 965,
 304         'Uuml'     => 220,
 305         'uuml'     => 252,
 306         'weierp'   => 8472,
 307         'Xi'       => 926,
 308         'xi'       => 958,
 309         'Yacute'   => 221,
 310         'yacute'   => 253,
 311         'yen'      => 165,
 312         'Yuml'     => 376,
 313         'yuml'     => 255,
 314         'Zeta'     => 918,
 315         'zeta'     => 950,
 316         'zwj'      => 8205,
 317         'zwnj'     => 8204 );
 318
 319 /**
 320  * Character entity aliases accepted by MediaWiki
 321  */
 322 global $wgHtmlEntityAliases;
 323 $wgHtmlEntityAliases = array(
 324         'רלמ' => 'rlm',
 325         'رلم' => 'rlm',
 326 );
 327
 328
 329 /**
 330  * XHTML sanitizer for MediaWiki
 331  * @ingroup Parser
 332  */
 333 class Sanitizer {
 334         const NONE = 0;
 335         const INITIAL_NONLETTER = 1;
 336
 337         /**
 338          * Cleans up HTML, removes dangerous tags and attributes, and
 339          * removes HTML comments
 340          * @private
 341          * @param string $text
 342          * @param callback $processCallback to do any variable or parameter replacements in HTML attribute values
 343          * @param array $args for the processing callback
 344          * @return string
 345          */
 346         static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array() ) {
 347                 global $wgUseTidy;
 348
 349                 static $htmlpairs, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 350                         $htmllist, $listtags, $htmlsingleallowed, $htmlelements, $staticInitialised;
 351
 352                 wfProfileIn( __METHOD__ );
 353
 354                 if ( !$staticInitialised ) {
 355
 356                         $htmlpairs = array_merge( $extratags, array( # Tags that must be closed
 357                                 'b', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 358                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 359                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 360                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 361                                 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'u'
 362                         ) );
 363                         $htmlsingle = array(
 364                                 'br', 'hr', 'li', 'dt', 'dd'
 365                         );
 366                         $htmlsingleonly = array( # Elements that cannot have close tags
 367                                 'br', 'hr'
 368                         );
 369                         $htmlnest = array( # Tags that can be nested--??
 370                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 371                                 'dl', 'font', 'big', 'small', 'sub', 'sup', 'span'
 372                         );
 373                         $tabletags = array( # Can only appear inside table, we will close them
 374                                 'td', 'th', 'tr',
 375                         );
 376                         $htmllist = array( # Tags used by list
 377                                 'ul','ol',
 378                         );
 379                         $listtags = array( # Tags that can appear in a list
 380                                 'li',
 381                         );
 382
 383                         $htmlsingleallowed = array_merge( $htmlsingle, $tabletags );
 384                         $htmlelements = array_merge( $htmlsingle, $htmlpairs, $htmlnest );
 385
 386                         # Convert them all to hashtables for faster lookup
 387                         $vars = array( 'htmlpairs', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 388                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelements' );
 389                         foreach ( $vars as $var ) {
 390                                 $$var = array_flip( $$var );
 391                         }
 392                         $staticInitialised = true;
 393                 }
 394
 395                 # Remove HTML comments
 396                 $text = Sanitizer::removeHTMLcomments( $text );
 397                 $bits = explode( '<', $text );
 398                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 399                 if(!$wgUseTidy) {
 400                         $tagstack = $tablestack = array();
 401                         foreach ( $bits as $x ) {
 402                                 $regs = array();
 403                                 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 404                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 405                                 } else {
 406                                         $slash = $t = $params = $brace = $rest = null;
 407                                 }
 408
 409                                 $badtag = 0 ;
 410                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 411                                         # Check our stack
 412                                         if ( $slash ) {
 413                                                 # Closing a tag...
 414                                                 if( isset( $htmlsingleonly[$t] ) ) {
 415                                                         $badtag = 1;
 416                                                 } elseif ( ( $ot = @array_pop( $tagstack ) ) != $t ) {
 417                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 418                                                                 # Pop all elements with an optional close tag
 419                                                                 # and see if we find a match below them
 420                                                                 $optstack = array();
 421                                                                 array_push ($optstack, $ot);
 422                                                                 while ( ( ( $ot = @array_pop( $tagstack ) ) != $t ) &&
 423                                                                                 isset( $htmlsingleallowed[$ot] ) )
 424                                                                 {
 425                                                                         array_push ($optstack, $ot);
 426                                                                 }
 427                                                                 if ( $t != $ot ) {
 428                                                                         # No match. Push the optinal elements back again
 429                                                                         $badtag = 1;
 430                                                                         while ( $ot = @array_pop( $optstack ) ) {
 431                                                                                 array_push( $tagstack, $ot );
 432                                                                         }
 433                                                                 }
 434                                                         } else {
 435                                                                 @array_push( $tagstack, $ot );
 436                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 437                                                                 if(!(isset( $htmllist[$ot] ) && isset( $listtags[$t] ) )) {
 438                                                                         $badtag = 1;
 439                                                                 }
 440                                                         }
 441                                                 } else {
 442                                                         if ( $t == 'table' ) {
 443                                                                 $tagstack = array_pop( $tablestack );
 444                                                         }
 445                                                 }
 446                                                 $newparams = '';
 447                                         } else {
 448                                                 # Keep track for later
 449                                                 if ( isset( $tabletags[$t] ) &&
 450                                                 ! in_array( 'table', $tagstack ) ) {
 451                                                         $badtag = 1;
 452                                                 } else if ( in_array( $t, $tagstack ) &&
 453                                                 ! isset( $htmlnest [$t ] ) ) {
 454                                                         $badtag = 1 ;
 455                                                 # Is it a self closed htmlpair ? (bug 5487)
 456                                                 } else if( $brace == '/>' &&
 457                                                 isset( $htmlpairs[$t] ) ) {
 458                                                         $badtag = 1;
 459                                                 } elseif( isset( $htmlsingleonly[$t] ) ) {
 460                                                         # Hack to force empty tag for uncloseable elements
 461                                                         $brace = '/>';
 462                                                 } else if( isset( $htmlsingle[$t] ) ) {
 463                                                         # Hack to not close $htmlsingle tags
 464                                                         $brace = NULL;
 465                                                 } else if( isset( $tabletags[$t] )
 466                                                 &&  in_array($t ,$tagstack) ) {
 467                                                         // New table tag but forgot to close the previous one
 468                                                         $text .= "</$t>";
 469                                                 } else {
 470                                                         if ( $t == 'table' ) {
 471                                                                 array_push( $tablestack, $tagstack );
 472                                                                 $tagstack = array();
 473                                                         }
 474                                                         array_push( $tagstack, $t );
 475                                                 }
 476
 477                                                 # Replace any variables or template parameters with
 478                                                 # plaintext results.
 479                                                 if( is_callable( $processCallback ) ) {
 480                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 481                                                 }
 482
 483                                                 # Strip non-approved attributes from the tag
 484                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 485                                         }
 486                                         if ( ! $badtag ) {
 487                                                 $rest = str_replace( '>', '&gt;', $rest );
 488                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 489                                                 $text .= "<$slash$t$newparams$close>$rest";
 490                                                 continue;
 491                                         }
 492                                 }
 493                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 494                         }
 495                         # Close off any remaining tags
 496                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 497                                 $text .= "</$t>\n";
 498                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 499                         }
 500                 } else {
 501                         # this might be possible using tidy itself
 502                         foreach ( $bits as $x ) {
 503                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 504                                 $x, $regs );
 505                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 506                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 507                                         if( is_callable( $processCallback ) ) {
 508                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 509                                         }
 510                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 511                                         $rest = str_replace( '>', '&gt;', $rest );
 512                                         $text .= "<$slash$t$newparams$brace$rest";
 513                                 } else {
 514                                         $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 515                                 }
 516                         }
 517                 }
 518                 wfProfileOut( __METHOD__ );
 519                 return $text;
 520         }
 521
 522         /**
 523          * Remove '<!--', '-->', and everything between.
 524          * To avoid leaving blank lines, when a comment is both preceded
 525          * and followed by a newline (ignoring spaces), trim leading and
 526          * trailing spaces and one of the newlines.
 527          *
 528          * @private
 529          * @param string $text
 530          * @return string
 531          */
 532         static function removeHTMLcomments( $text ) {
 533                 wfProfileIn( __METHOD__ );
 534                 while (($start = strpos($text, '<!--')) !== false) {
 535                         $end = strpos($text, '-->', $start + 4);
 536                         if ($end === false) {
 537                                 # Unterminated comment; bail out
 538                                 break;
 539                         }
 540
 541                         $end += 3;
 542
 543                         # Trim space and newline if the comment is both
 544                         # preceded and followed by a newline
 545                         $spaceStart = max($start - 1, 0);
 546                         $spaceLen = $end - $spaceStart;
 547                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 548                                 $spaceStart--;
 549                                 $spaceLen++;
 550                         }
 551                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 552                                 $spaceLen++;
 553                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 554                                 # Remove the comment, leading and trailing
 555                                 # spaces, and leave only one newline.
 556                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 557                         }
 558                         else {
 559                                 # Remove just the comment.
 560                                 $text = substr_replace($text, '', $start, $end - $start);
 561                         }
 562                 }
 563                 wfProfileOut( __METHOD__ );
 564                 return $text;
 565         }
 566
 567         /**
 568          * Take an array of attribute names and values and normalize or discard
 569          * illegal values for the given element type.
 570          *
 571          * - Discards attributes not on a whitelist for the given element
 572          * - Unsafe style attributes are discarded
 573          * - Invalid id attributes are reencoded
 574          *
 575          * @param array $attribs
 576          * @param string $element
 577          * @return array
 578          *
 579          * @todo Check for legal values where the DTD limits things.
 580          * @todo Check for unique id attribute :P
 581          */
 582         static function validateTagAttributes( $attribs, $element ) {
 583                 return Sanitizer::validateAttributes( $attribs,
 584                         Sanitizer::attributeWhitelist( $element ) );
 585         }
 586
 587         /**
 588          * Take an array of attribute names and values and normalize or discard
 589          * illegal values for the given whitelist.
 590          *
 591          * - Discards attributes not the given whitelist
 592          * - Unsafe style attributes are discarded
 593          * - Invalid id attributes are reencoded
 594          *
 595          * @param array $attribs
 596          * @param array $whitelist list of allowed attribute names
 597          * @return array
 598          *
 599          * @todo Check for legal values where the DTD limits things.
 600          * @todo Check for unique id attribute :P
 601          */
 602         static function validateAttributes( $attribs, $whitelist ) {
 603                 $whitelist = array_flip( $whitelist );
 604                 $out = array();
 605                 foreach( $attribs as $attribute => $value ) {
 606                         if( !isset( $whitelist[$attribute] ) ) {
 607                                 continue;
 608                         }
 609                         # Strip javascript "expression" from stylesheets.
 610                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 611                         if( $attribute == 'style' ) {
 612                                 $value = Sanitizer::checkCss( $value );
 613                                 if( $value === false ) {
 614                                         # haxx0r
 615                                         continue;
 616                                 }
 617                         }
 618
 619                         if ( $attribute === 'id' )
 620                                 $value = Sanitizer::escapeId( $value );
 621
 622                         // If this attribute was previously set, override it.
 623                         // Output should only have one attribute of each name.
 624                         $out[$attribute] = $value;
 625                 }
 626                 return $out;
 627         }
 628
 629         /**
 630          * Merge two sets of HTML attributes.
 631          * Conflicting items in the second set will override those
 632          * in the first, except for 'class' attributes which will be
 633          * combined.
 634          *
 635          * @todo implement merging for other attributes such as style
 636          * @param array $a
 637          * @param array $b
 638          * @return array
 639          */
 640         static function mergeAttributes( $a, $b ) {
 641                 $out = array_merge( $a, $b );
 642                 if( isset( $a['class'] )
 643                         && isset( $b['class'] )
 644                         && $a['class'] !== $b['class'] ) {
 645
 646                         $out['class'] = implode( ' ',
 647                                 array_unique(
 648                                         preg_split( '/\s+/',
 649                                                 $a['class'] . ' ' . $b['class'],
 650                                                 -1,
 651                                                 PREG_SPLIT_NO_EMPTY ) ) );
 652                 }
 653                 return $out;
 654         }
 655
 656         /**
 657          * Pick apart some CSS and check it for forbidden or unsafe structures.
 658          * Returns a sanitized string, or false if it was just too evil.
 659          *
 660          * Currently URL references, 'expression', 'tps' are forbidden.
 661          *
 662          * @param string $value
 663          * @return mixed
 664          */
 665         static function checkCss( $value ) {
 666                 $stripped = Sanitizer::decodeCharReferences( $value );
 667
 668                 // Remove any comments; IE gets token splitting wrong
 669                 $stripped = StringUtils::delimiterReplace( '/*', '*/', ' ', $stripped );
 670
 671                 $value = $stripped;
 672
 673                 // ... and continue checks
 674                 $stripped = preg_replace( '!\\\\([0-9A-Fa-f]{1,6})[ \\n\\r\\t\\f]?!e',
 675                         'codepointToUtf8(hexdec("$1"))', $stripped );
 676                 $stripped = str_replace( '\\', '', $stripped );
 677                 if( preg_match( '/(?:expression|tps*:\/\/|url\\s*\().*/is',
 678                                 $stripped ) ) {
 679                         # haxx0r
 680                         return false;
 681                 }
 682
 683                 return $value;
 684         }
 685
 686         /**
 687          * Take a tag soup fragment listing an HTML element's attributes
 688          * and normalize it to well-formed XML, discarding unwanted attributes.
 689          * Output is safe for further wikitext processing, with escaping of
 690          * values that could trigger problems.
 691          *
 692          * - Normalizes attribute names to lowercase
 693          * - Discards attributes not on a whitelist for the given element
 694          * - Turns broken or invalid entities into plaintext
 695          * - Double-quotes all attribute values
 696          * - Attributes without values are given the name as attribute
 697          * - Double attributes are discarded
 698          * - Unsafe style attributes are discarded
 699          * - Prepends space if there are attributes.
 700          *
 701          * @param string $text
 702          * @param string $element
 703          * @return string
 704          */
 705         static function fixTagAttributes( $text, $element ) {
 706                 if( trim( $text ) == '' ) {
 707                         return '';
 708                 }
 709
 710                 $stripped = Sanitizer::validateTagAttributes(
 711                         Sanitizer::decodeTagAttributes( $text ), $element );
 712
 713                 $attribs = array();
 714                 foreach( $stripped as $attribute => $value ) {
 715                         $encAttribute = htmlspecialchars( $attribute );
 716                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 717
 718                         $attribs[] = "$encAttribute=\"$encValue\"";
 719                 }
 720                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 721         }
 722
 723         /**
 724          * Encode an attribute value for HTML output.
 725          * @param $text
 726          * @return HTML-encoded text fragment
 727          */
 728         static function encodeAttribute( $text ) {
 729                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 730
 731                 // Whitespace is normalized during attribute decoding,
 732                 // so if we've been passed non-spaces we must encode them
 733                 // ahead of time or they won't be preserved.
 734                 $encValue = strtr( $encValue, array(
 735                         "\n" => '&#10;',
 736                         "\r" => '&#13;',
 737                         "\t" => '&#9;',
 738                 ) );
 739
 740                 return $encValue;
 741         }
 742
 743         /**
 744          * Encode an attribute value for HTML tags, with extra armoring
 745          * against further wiki processing.
 746          * @param $text
 747          * @return HTML-encoded text fragment
 748          */
 749         static function safeEncodeAttribute( $text ) {
 750                 $encValue = Sanitizer::encodeAttribute( $text );
 751
 752                 # Templates and links may be expanded in later parsing,
 753                 # creating invalid or dangerous output. Suppress this.
 754                 $encValue = strtr( $encValue, array(
 755                         '<'    => '&lt;',   // This should never happen,
 756                         '>'    => '&gt;',   // we've received invalid input
 757                         '"'    => '&quot;', // which should have been escaped.
 758                         '{'    => '&#123;',
 759                         '['    => '&#91;',
 760                         "''"   => '&#39;&#39;',
 761                         'ISBN' => '&#73;SBN',
 762                         'RFC'  => '&#82;FC',
 763                         'PMID' => '&#80;MID',
 764                         '|'    => '&#124;',
 765                         '__'   => '&#95;_',
 766                 ) );
 767
 768                 # Stupid hack
 769                 $encValue = preg_replace_callback(
 770                         '/(' . wfUrlProtocols() . ')/',
 771                         array( 'Sanitizer', 'armorLinksCallback' ),
 772                         $encValue );
 773                 return $encValue;
 774         }
 775
 776         /**
 777          * Given a value escape it so that it can be used in an id attribute and
 778          * return it, this does not validate the value however (see first link)
 779          *
 780          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
 781          *                                                          in the id and
 782          *                                                          name attributes
 783          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
 784          *
 785          * @param string $id    Id to validate
 786          * @param int    $flags Currently only two values: Sanitizer::INITIAL_NONLETTER
 787          *                      (default) permits initial non-letter characters,
 788          *                      such as if you're adding a prefix to them.
 789          *                      Sanitizer::NONE will prepend an 'x' if the id
 790          *                      would otherwise start with a nonletter.
 791          * @return string
 792          */
 793         static function escapeId( $id, $flags = Sanitizer::INITIAL_NONLETTER ) {
 794                 static $replace = array(
 795                         '%3A' => ':',
 796                         '%' => '.'
 797                 );
 798
 799                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
 800                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
 801
 802                 if( ~$flags & Sanitizer::INITIAL_NONLETTER
 803                 && !preg_match( '/[a-zA-Z]/', $id[0] ) ) {
 804                         // Initial character must be a letter!
 805                         $id = "x$id";
 806                 }
 807                 return $id;
 808         }
 809
 810         /**
 811          * Given a value, escape it so that it can be used as a CSS class and
 812          * return it.
 813          *
 814          * @todo For extra validity, input should be validated UTF-8.
 815          *
 816          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
 817          *
 818          * @param string $class
 819          * @return string
 820          */
 821         static function escapeClass( $class ) {
 822                 // Convert ugly stuff to underscores and kill underscores in ugly places
 823                 return rtrim(preg_replace(
 824                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
 825                         '_',
 826                         $class ), '_');
 827         }
 828
 829         /**
 830          * Regex replace callback for armoring links against further processing.
 831          * @param array $matches
 832          * @return string
 833          * @private
 834          */
 835         private static function armorLinksCallback( $matches ) {
 836                 return str_replace( ':', '&#58;', $matches[1] );
 837         }
 838
 839         /**
 840          * Return an associative array of attribute names and values from
 841          * a partial tag string. Attribute names are forces to lowercase,
 842          * character references are decoded to UTF-8 text.
 843          *
 844          * @param string
 845          * @return array
 846          */
 847         static function decodeTagAttributes( $text ) {
 848                 $attribs = array();
 849
 850                 if( trim( $text ) == '' ) {
 851                         return $attribs;
 852                 }
 853
 854                 $pairs = array();
 855                 if( !preg_match_all(
 856                         MW_ATTRIBS_REGEX,
 857                         $text,
 858                         $pairs,
 859                         PREG_SET_ORDER ) ) {
 860                         return $attribs;
 861                 }
 862
 863                 foreach( $pairs as $set ) {
 864                         $attribute = strtolower( $set[1] );
 865                         $value = Sanitizer::getTagAttributeCallback( $set );
 866
 867                         // Normalize whitespace
 868                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
 869                         $value = trim( $value );
 870
 871                         // Decode character references
 872                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
 873                 }
 874                 return $attribs;
 875         }
 876
 877         /**
 878          * Pick the appropriate attribute value from a match set from the
 879          * MW_ATTRIBS_REGEX matches.
 880          *
 881          * @param array $set
 882          * @return string
 883          * @private
 884          */
 885         private static function getTagAttributeCallback( $set ) {
 886                 if( isset( $set[6] ) ) {
 887                         # Illegal #XXXXXX color with no quotes.
 888                         return $set[6];
 889                 } elseif( isset( $set[5] ) ) {
 890                         # No quotes.
 891                         return $set[5];
 892                 } elseif( isset( $set[4] ) ) {
 893                         # Single-quoted
 894                         return $set[4];
 895                 } elseif( isset( $set[3] ) ) {
 896                         # Double-quoted
 897                         return $set[3];
 898                 } elseif( !isset( $set[2] ) ) {
 899                         # In XHTML, attributes must have a value.
 900                         # For 'reduced' form, return explicitly the attribute name here.
 901                         return $set[1];
 902                 } else {
 903                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
 904                 }
 905         }
 906
 907         /**
 908          * Normalize whitespace and character references in an XML source-
 909          * encoded text for an attribute value.
 910          *
 911          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
 912          * but note that we're not returning the value, but are returning
 913          * XML source fragments that will be slapped into output.
 914          *
 915          * @param string $text
 916          * @return string
 917          * @private
 918          */
 919         private static function normalizeAttributeValue( $text ) {
 920                 return str_replace( '"', '&quot;',
 921                         self::normalizeWhitespace(
 922                                 Sanitizer::normalizeCharReferences( $text ) ) );
 923         }
 924
 925         private static function normalizeWhitespace( $text ) {
 926                 return preg_replace(
 927                         '/\r\n|[\x20\x0d\x0a\x09]/',
 928                         ' ',
 929                         $text );
 930         }
 931
 932         /**
 933          * Ensure that any entities and character references are legal
 934          * for XML and XHTML specifically. Any stray bits will be
 935          * &amp;-escaped to result in a valid text fragment.
 936          *
 937          * a. any named char refs must be known in XHTML
 938          * b. any numeric char refs must be legal chars, not invalid or forbidden
 939          * c. use &#x, not &#X
 940          * d. fix or reject non-valid attributes
 941          *
 942          * @param string $text
 943          * @return string
 944          * @private
 945          */
 946         static function normalizeCharReferences( $text ) {
 947                 return preg_replace_callback(
 948                         MW_CHAR_REFS_REGEX,
 949                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
 950                         $text );
 951         }
 952         /**
 953          * @param string $matches
 954          * @return string
 955          */
 956         static function normalizeCharReferencesCallback( $matches ) {
 957                 $ret = null;
 958                 if( $matches[1] != '' ) {
 959                         $ret = Sanitizer::normalizeEntity( $matches[1] );
 960                 } elseif( $matches[2] != '' ) {
 961                         $ret = Sanitizer::decCharReference( $matches[2] );
 962                 } elseif( $matches[3] != ''  ) {
 963                         $ret = Sanitizer::hexCharReference( $matches[3] );
 964                 } elseif( $matches[4] != '' ) {
 965                         $ret = Sanitizer::hexCharReference( $matches[4] );
 966                 }
 967                 if( is_null( $ret ) ) {
 968                         return htmlspecialchars( $matches[0] );
 969                 } else {
 970                         return $ret;
 971                 }
 972         }
 973
 974         /**
 975          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
 976          * return the named entity reference as is. If the entity is a
 977          * MediaWiki-specific alias, returns the HTML equivalent. Otherwise,
 978          * returns HTML-escaped text of pseudo-entity source (eg &amp;foo;)
 979          *
 980          * @param string $name
 981          * @return string
 982          * @static
 983          */
 984         static function normalizeEntity( $name ) {
 985                 global $wgHtmlEntities, $wgHtmlEntityAliases;
 986                 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
 987                         return "&{$wgHtmlEntityAliases[$name]};";
 988                 } elseif( isset( $wgHtmlEntities[$name] ) ) {
 989                         return "&$name;";
 990                 } else {
 991                         return "&amp;$name;";
 992                 }
 993         }
 994
 995         static function decCharReference( $codepoint ) {
 996                 $point = intval( $codepoint );
 997                 if( Sanitizer::validateCodepoint( $point ) ) {
 998                         return sprintf( '&#%d;', $point );
 999                 } else {
1000                         return null;
1001                 }
1002         }
1003
1004         static function hexCharReference( $codepoint ) {
1005                 $point = hexdec( $codepoint );
1006                 if( Sanitizer::validateCodepoint( $point ) ) {
1007                         return sprintf( '&#x%x;', $point );
1008                 } else {
1009                         return null;
1010                 }
1011         }
1012
1013         /**
1014          * Returns true if a given Unicode codepoint is a valid character in XML.
1015          * @param int $codepoint
1016          * @return bool
1017          */
1018         private static function validateCodepoint( $codepoint ) {
1019                 return ($codepoint ==    0x09)
1020                         || ($codepoint ==    0x0a)
1021                         || ($codepoint ==    0x0d)
1022                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1023                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1024                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1025         }
1026
1027         /**
1028          * Decode any character references, numeric or named entities,
1029          * in the text and return a UTF-8 string.
1030          *
1031          * @param string $text
1032          * @return string
1033          * @public
1034          * @static
1035          */
1036         public static function decodeCharReferences( $text ) {
1037                 return preg_replace_callback(
1038                         MW_CHAR_REFS_REGEX,
1039                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1040                         $text );
1041         }
1042
1043         /**
1044          * @param string $matches
1045          * @return string
1046          */
1047         static function decodeCharReferencesCallback( $matches ) {
1048                 if( $matches[1] != '' ) {
1049                         return Sanitizer::decodeEntity( $matches[1] );
1050                 } elseif( $matches[2] != '' ) {
1051                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1052                 } elseif( $matches[3] != ''  ) {
1053                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1054                 } elseif( $matches[4] != '' ) {
1055                         return  Sanitizer::decodeChar( hexdec( $matches[4] ) );
1056                 }
1057                 # Last case should be an ampersand by itself
1058                 return $matches[0];
1059         }
1060
1061         /**
1062          * Return UTF-8 string for a codepoint if that is a valid
1063          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1064          * @param int $codepoint
1065          * @return string
1066          * @private
1067          */
1068         static function decodeChar( $codepoint ) {
1069                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1070                         return codepointToUtf8( $codepoint );
1071                 } else {
1072                         return UTF8_REPLACEMENT;
1073                 }
1074         }
1075
1076         /**
1077          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1078          * return the UTF-8 encoding of that character. Otherwise, returns
1079          * pseudo-entity source (eg &foo;)
1080          *
1081          * @param string $name
1082          * @return string
1083          */
1084         static function decodeEntity( $name ) {
1085                 global $wgHtmlEntities, $wgHtmlEntityAliases;
1086                 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
1087                         $name = $wgHtmlEntityAliases[$name];
1088                 }
1089                 if( isset( $wgHtmlEntities[$name] ) ) {
1090                         return codepointToUtf8( $wgHtmlEntities[$name] );
1091                 } else {
1092                         return "&$name;";
1093                 }
1094         }
1095
1096         /**
1097          * Fetch the whitelist of acceptable attributes for a given
1098          * element name.
1099          *
1100          * @param string $element
1101          * @return array
1102          */
1103         static function attributeWhitelist( $element ) {
1104                 static $list;
1105                 if( !isset( $list ) ) {
1106                         $list = Sanitizer::setupAttributeWhitelist();
1107                 }
1108                 return isset( $list[$element] )
1109                         ? $list[$element]
1110                         : array();
1111         }
1112
1113         /**
1114          * @todo Document it a bit
1115          * @return array
1116          */
1117         static function setupAttributeWhitelist() {
1118                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1119                 $block = array_merge( $common, array( 'align' ) );
1120                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1121                 $tablecell = array( 'abbr',
1122                                     'axis',
1123                                     'headers',
1124                                     'scope',
1125                                     'rowspan',
1126                                     'colspan',
1127                                     'nowrap', # deprecated
1128                                     'width',  # deprecated
1129                                     'height', # deprecated
1130                                     'bgcolor' # deprecated
1131                                     );
1132
1133                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1134                 # See: http://www.w3.org/TR/html4/
1135                 $whitelist = array (
1136                         # 7.5.4
1137                         'div'        => $block,
1138                         'center'     => $common, # deprecated
1139                         'span'       => $block, # ??
1140
1141                         # 7.5.5
1142                         'h1'         => $block,
1143                         'h2'         => $block,
1144                         'h3'         => $block,
1145                         'h4'         => $block,
1146                         'h5'         => $block,
1147                         'h6'         => $block,
1148
1149                         # 7.5.6
1150                         # address
1151
1152                         # 8.2.4
1153                         # bdo
1154
1155                         # 9.2.1
1156                         'em'         => $common,
1157                         'strong'     => $common,
1158                         'cite'       => $common,
1159                         # dfn
1160                         'code'       => $common,
1161                         # samp
1162                         # kbd
1163                         'var'        => $common,
1164                         # abbr
1165                         # acronym
1166
1167                         # 9.2.2
1168                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1169                         # q
1170
1171                         # 9.2.3
1172                         'sub'        => $common,
1173                         'sup'        => $common,
1174
1175                         # 9.3.1
1176                         'p'          => $block,
1177
1178                         # 9.3.2
1179                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1180
1181                         # 9.3.4
1182                         'pre'        => array_merge( $common, array( 'width' ) ),
1183
1184                         # 9.4
1185                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1186                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1187
1188                         # 10.2
1189                         'ul'         => array_merge( $common, array( 'type' ) ),
1190                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1191                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1192
1193                         # 10.3
1194                         'dl'         => $common,
1195                         'dd'         => $common,
1196                         'dt'         => $common,
1197
1198                         # 11.2.1
1199                         'table'      => array_merge( $common,
1200                                                                 array( 'summary', 'width', 'border', 'frame',
1201                                                                                 'rules', 'cellspacing', 'cellpadding',
1202                                                                                 'align', 'bgcolor',
1203                                                                 ) ),
1204
1205                         # 11.2.2
1206                         'caption'    => array_merge( $common, array( 'align' ) ),
1207
1208                         # 11.2.3
1209                         'thead'      => array_merge( $common, $tablealign ),
1210                         'tfoot'      => array_merge( $common, $tablealign ),
1211                         'tbody'      => array_merge( $common, $tablealign ),
1212
1213                         # 11.2.4
1214                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1215                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1216
1217                         # 11.2.5
1218                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1219
1220                         # 11.2.6
1221                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1222                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1223
1224                         # 13.2
1225                         # Not usually allowed, but may be used for extension-style hooks
1226                         # such as <math> when it is rasterized
1227                         'img'        => array_merge( $common, array( 'alt' ) ),
1228
1229                         # 15.2.1
1230                         'tt'         => $common,
1231                         'b'          => $common,
1232                         'i'          => $common,
1233                         'big'        => $common,
1234                         'small'      => $common,
1235                         'strike'     => $common,
1236                         's'          => $common,
1237                         'u'          => $common,
1238
1239                         # 15.2.2
1240                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1241                         # basefont
1242
1243                         # 15.3
1244                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1245
1246                         # XHTML Ruby annotation text module, simple ruby only.
1247                         # http://www.w3c.org/TR/ruby/
1248                         'ruby'       => $common,
1249                         # rbc
1250                         # rtc
1251                         'rb'         => $common,
1252                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1253                         'rp'         => $common,
1254
1255                         # MathML root element, where used for extensions
1256                         # 'title' may not be 100% valid here; it's XHTML
1257                         # http://www.w3.org/TR/REC-MathML/
1258                         'math'       => array( 'class', 'style', 'id', 'title' ),
1259                         );
1260                 return $whitelist;
1261         }
1262
1263         /**
1264          * Take a fragment of (potentially invalid) HTML and return
1265          * a version with any tags removed, encoded as plain text.
1266          *
1267          * Warning: this return value must be further escaped for literal
1268          * inclusion in HTML output as of 1.10!
1269          *
1270          * @param string $text HTML fragment
1271          * @return string
1272          */
1273         static function stripAllTags( $text ) {
1274                 # Actual <tags>
1275                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1276
1277                 # Normalize &entities and whitespace
1278                 $text = self::decodeCharReferences( $text );
1279                 $text = self::normalizeWhitespace( $text );
1280
1281                 return $text;
1282         }
1283
1284         /**
1285          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1286          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1287          * PHP 5.1 doesn't.
1288          *
1289          * Use for passing XHTML fragments to PHP's XML parsing functions
1290          *
1291          * @return string
1292          * @static
1293          */
1294         static function hackDocType() {
1295                 global $wgHtmlEntities;
1296                 $out = "<!DOCTYPE html [\n";
1297                 foreach( $wgHtmlEntities as $entity => $codepoint ) {
1298                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1299                 }
1300                 $out .= "]>\n";
1301                 return $out;
1302         }
1303
1304         static function cleanUrl( $url, $hostname=true ) {
1305                 # Normalize any HTML entities in input. They will be
1306                 # re-escaped by makeExternalLink().
1307                 $url = Sanitizer::decodeCharReferences( $url );
1308
1309                 # Escape any control characters introduced by the above step
1310                 $url = preg_replace( '/[\][<>"\\x00-\\x20\\x7F]/e', "urlencode('\\0')", $url );
1311
1312                 # Validate hostname portion
1313                 $matches = array();
1314                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1315                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1316
1317                         // Characters that will be ignored in IDNs.
1318                         // http://tools.ietf.org/html/3454#section-3.1
1319                         // Strip them before further processing so blacklists and such work.
1320                         $strip = "/
1321                                 \\s|          # general whitespace
1322                                 \xc2\xad|     # 00ad SOFT HYPHEN
1323                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1324                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1325                                 \xe2\x81\xa0| # 2060 WORD JOINER
1326                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1327                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1328                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1329                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1330                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1331                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1332                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1333                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe00f VARIATION SELECTOR-1-16
1334                                 /xuD";
1335
1336                         $host = preg_replace( $strip, '', $host );
1337
1338                         // @fixme: validate hostnames here
1339
1340                         return $protocol . $host . $rest;
1341                 } else {
1342                         return $url;
1343                 }
1344         }
1345
1346 }